mg の新機能

変更したのはもう去年のことだが、今使っている mg には、これまでとまったく違う検索機能が追加されている。

mg に限らず grep 系のコマンドでは、検索エンジンのような指定がしにくい。たとえば foo と bar の両方が含まれている行を検索しようとすると案外面倒なのである。普通はパイプを使ってこんな風にするのだろうか。

% grep foo file | grep bar

mg で1回でこれを実行するとすれば、

% mg -e 'foo.*bar|bar.*foo'

という具合か。でも、文字列の数が増えてくるとパターンが極めて複雑になってしまう。

先読みを使うとこんな風にも書ける。

% mg -e '(?=.*foo)(?=.*bar)'

これはこれで、なかなかご機嫌な使い方だ。
さらに buz を含まない行を探したければこう書けばいい。

% mg -e '(?=.*foo)(?=.*bar)(?!.*buz)'

というわけで、それをもっと簡単に指定できる機能を実現してみたのである。
最初は必要があって追加した機能だが、その後なかなか使う機会がなくて、リリースに至らずにいる。

EXTENDED PATTERN SEARCH

       Mg  now supports completely new search method.  This feature is invoked

       by -x flag, and then the specified pattern is interpreted in  the  dif-

       ferent manner.
       Pattern  string is treated as a colleciton of tokens separated by white

       space.  Each component will be searched  independently,  but  only  the

       line which contains all of them will be printed.  For examle,
            mg -xp 'foo bar buz' ...
       will print lines which contain all of `foo', `bar' and `buz'.  They can

       be found in any order and/or any place in the string.  So this  command

       find all of following texts.
            foo bar buz

            buz bar foo

            the foo, bar and buz
       If  you  want  to  use  OR  syntax, prepend question (`?') sign on each

       token, or use regular expression in pattern:
            mg -xp 'foo bar buz ?yabba ?dabba ?doo'

            mg -exp 'foo bar buz yabba|dabba|doo'
       This command will print the line which contains all of `foo', `bar' and

       `buz'  and  one  or more from `yabba', `dabba' or `doo'.  Note that you

       need to use -e option to enable regular expression interpretation.
       Please be aware that multiple `?' preceded tokens are treated all mixed

       together.   That means `?A|B ?C|D' is equivalent to `?A|B|C|D'.  If you

       want to mean `(A or B) and (C or D)', use AND  syntax  instead  of  OR:

       `A|B C|D'.
       NOT  operator  can  be  specified by prefixing the token by minus (`-')

       sign.  Next example will show the line which  contain  both  `foo'  and

       `bar' but none of `yabba' or `dabba' or `doo'.
            mg -xp 'foo bar -yabba -dabba -doo'

            mg -exp 'foo bar -yabba|dabba|doo'
       It  is  ok to set plus (`+') sign before positive AND token, but it has

       no effect.
       When executed with -o option, the paragraph which  contains  all  these

       components will be found.  This style is much more usefull, actually.
       You  can't  use  double quote to include white space within each token.

       Separate options are prepared to spcify  each  component  individually;

       --and, --or and --not.  These options can be used multiple times.
            mg --and 'foo bar' --and buz --or 'yabba dabba' --or doo ...
       Long  option --xp is equivalent to the combination of -x and (optional)

       -p options.  Author decided to override single character option -x,  to

       make it posssible using in this way:
            mg -oeiQxp 'foo bar buz' ...